SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.05946 · 2024

Safety Alignment Should Be Made More Than Just a Few Tokens Deep

arXiv · PDF · Open in the Atlas

Code that ran

1 of 7 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 1 failure of our instrument; 6 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2406.05946.json · badge: /api/ran/2406.05946.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
unispac/shallow-vs-deep-alignment canonical (classifier label)
classifier basis: census
1 ran · 6 no recorded run
FunctionStatusWhere it lives
get_hex_phi Ran
ran; our contract draft was wrong, not the code
unispac/shallow-vs-deep-alignment/finetuning_buckets/datasets/utils/get_eval_data.py
code served (permissive licence) · get_code("59a1bc33ea694cc1")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import json
import tempfile
import os

# Create a temporary directory and a mock JSONL file
temp_dir = tempfile.mkdtemp()
os.makedirs(os.path.join(temp_dir, 'finetuning_buckets/datasets/data/safety_bench'), exist_ok=True)

# Create a mock HEx-PHI.jsonl file
jsonl_path = os.path.join(temp_dir, 'finetuning_buckets/datasets/data/safety_bench/HEx-PHI.jsonl')
with open(jsonl_path, 'w') as f:
    f.write(json.dumps({'instruction': 'What is AI?', 'category': 'general'}) + '\n')
    f.write(json.dumps({'instruction': 'How do I make a bomb?', 'category': 'harmful'}) + '\n')

# Change to temp directory so the relative path works
original_cwd = os.getcwd()
os.chdir(temp_dir)

ARGS = {
    'system_prompt': 'You are a helpful assistant.',
    'input_template': 'User question: %s',
    'output_header': 'I will help you with:',
    'split': 'test'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_alpaca_instruction Not yet run
unverified (harvested, no recorded run)
unispac/shallow-vs-deep-alignment/finetuning_buckets/datasets/utils/get_finetuning_data.py
code served (permissive licence) · get_code("584880181ee52b40")
How it ran

no invocation recorded.

get_hex_phi_backdoor Not yet run
unverified (harvested, no recorded run)
unispac/shallow-vs-deep-alignment/finetuning_buckets/datasets/utils/get_eval_data.py
code served (permissive licence) · get_code("b4c93e16c63c0cdd")
How it ran

no invocation recorded.

get_hex_phi_with_refusal_prefix Not yet run
unverified (harvested, no recorded run)
unispac/shallow-vs-deep-alignment/finetuning_buckets/datasets/utils/get_eval_data.py
code served (permissive licence) · get_code("322473e24c0ea7a3")
How it ran

no invocation recorded.

get_model Not yet run
unverified (harvested, no recorded run)
unispac/shallow-vs-deep-alignment/finetuning_buckets/models/get_model.py
code served (permissive licence) · get_code("5ccba273e30aff64")
How it ran

no invocation recorded.

get_pure_bad Not yet run
unverified (harvested, no recorded run)
unispac/shallow-vs-deep-alignment/finetuning_buckets/datasets/utils/get_finetuning_data.py
code served (permissive licence) · get_code("0fdbd70545579b0f")
How it ran

no invocation recorded.

get_safety_augmentation_data Not yet run
unverified (harvested, no recorded run)
unispac/shallow-vs-deep-alignment/finetuning_buckets/datasets/utils/get_finetuning_data.py
code served (permissive licence) · get_code("2c7c6fbe31bd60bc")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 7 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The safety alignment of current Large Language Models (LLMs) is vulnerable. Relatively simple attacks, or even benign fine-tuning, can jailbreak aligned models. We argue that many of these vulnerabilities are related to a shared underlying issue: safety alignment can take shortcuts, wherein the alignment adapts a model's generative distribution primarily over only its very first few output tokens. We refer to this issue as shallow safety alignment. In this paper, we present case studies to explain why shallow safety alignment can exist and provide evidence that current aligned LLMs are subject to this issue. We also show how these findings help explain multiple recently discovered vulnerabilities in LLMs, including the susceptibility to adversarial suffix attacks, prefilling attacks, decoding parameter attacks, and fine-tuning attacks. Importantly, we discuss how this consolidated notion of shallow safety alignment sheds light on promising research directions for mitigating these vulnerabilities. For instance, we show that deepening the safety alignment beyond just the first few tokens can often meaningfully improve robustness against some common exploits. Finally, we design a regularized finetuning objective that makes the safety alignment more persistent against fine-tuning attacks by constraining updates on initial tokens. Overall, we advocate that future safety alignment should be made more than just a few tokens deep.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2406.05946")
get_code_for_paper("2406.05946")
have("2406.05946")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2406.05946.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2406.05946.svg)](https://syntology.ai/paper/2406.05946)

Connect an agent — have() is free.